You don't need to transmit the audio. Just transcribe to text, remove any common or stop-words (a, an, the) and transmit what's left over in encrypted form to cover their tracks.